|Publication number||US6874001 B2|
|Application number||US 09/971,755|
|Publication date||Mar 29, 2005|
|Filing date||Oct 5, 2001|
|Priority date||Oct 5, 2001|
|Also published as||US20030069902, US20050187983|
|Publication number||09971755, 971755, US 6874001 B2, US 6874001B2, US-B2-6874001, US6874001 B2, US6874001B2|
|Inventors||Inderpal Singh Narang, Karen Wolfe Brannon, Suparna Bhattacharya, Hui-I Hsiao|
|Original Assignee||International Business Machines Corporation|
|Export Citation||BiBTeX, EndNote, RefMan|
|Patent Citations (6), Non-Patent Citations (2), Referenced by (75), Classifications (13), Legal Events (6)|
|External Links: USPTO, USPTO Assignment, Espacenet|
The present invention relates generally to electronic information technology and more particularly to electronic data management systems.
Content management systems store, access, and manage digital information or content in networked environments, including such information generated in e-business applications. Well integrated content has many applications, and for example, can accelerate business process automation across an enterprise for various sized e-businesses. Relevant content can include electronic documents, text files, XML and HTML files, digital audio and video files, scanned images, facsimiles, and the like.
Content management systems, especially those designed for managing distributed content, may also store meta-data describing an object or related information in a store that is separate from a file containing the object's content. DB2 Universal Database (UDB) is an example of a scalable database that can be used in content management systems as a database management system (DBMS). The DBMS is an essential part of the system for storing digital information or content. Meta-data can be stored in a repository of the DBMS. Examples of meta-data include information about data sources, access authorization, archive and backup histories, data accesses and content identification labels (e.g., video, audio, text).
A particular challenge arising in content management systems is maintaining consistency between file content and the associated meta-data from the point of view of an application accessing the content and meta-data (simply referred to hereinafter as a reader). If file and meta-data updates are tightly coupled (i.e. both updates happen within a single unified transaction), a transaction coordinator typically ensures a consistent view by locking out readers of meta-data as well as file data until the transaction is committed. Intermediate/uncommitted updates to either are not visible. However, this approach has a number of disadvantages including the circumstance that content edits can require considerable amounts of time. This approach also does not allow the convenience of directly accessing and updating the file content on the native filesystem, using native filesystem operations.
On the other hand, in systems where a loose transaction model is provided, and direct content edits are allowed, consistency between file-data and meta-data may not be guaranteed at all times.
A need therefore clearly exists for an improved technique for providing a consistent view of file data and meta-data in the presence of a loose-transaction model.
Management of External Data Using DATALINKS
Content can be referenced within a database using the DATALINK data type, which is part of the ANSI ISO standard described in document ISO/IEC 9075-9:2000(E), “Information Technology—Database Languages SQL—Part 9: Management of External Data (SQL/MED)”.
Datalinks (DL) is a mechanism that can be used with DB2 UDB, facilitating management of files residing outside of a database as though the files are logically within the database. Datalinks ensures referential integrity of the external files, provides access control for the files, and supports automatic and coordinated backup and restore capabilities. In this manner, a coordinated administration point is provided for file and database data. In the following description, a DATALINK type is referred to for SQL management of external data.
Datalinks maintains a reference to a file residing in a file system in a column of a DB2 table. The reference is stored as DATALINK data type, which uses a Uniform Resource Locator (URL). A table in the database may have one or more DATALINK columns. The database may also store meta-data about the files with DATALINK references in the same table or other tables. Thus, a row may contain a DATALINK reference and meta-data related to the external file. An SQL query or statement issued by an application or reader may be applied against the table on the meta-data, for example, to locate the (external) file of interest. The URL retrieved from the DATALINK column of the row is then used to access the external file via the native API of the file system or browser.
Further information regarding the datalink structure shown in
DataLinks supports two modes of “WRITE PERMISSION”:FS (File System) or BLOCKED. In the FS case, users are allowed to update a file while the file is linked to the database (i.e. the database has a DATALINK value which is currently pointing to the file). However, this mode does not provide file data recovery, which means if the disk crashes or a user needs to restore the database, there is no file backup data to recover from. This could cause inconsistency between the file data and the database data after RESTORE/ROLLFORWORD to a point in time other than the time of the crash. Moreover, WRITE PERMISSION FS does not support the SQL mediated access model like the one provided in READ PERMISSION DB.
On the other hand, WRITE PERMISSION BLOCKED provides data recovery for a file with a reference in a DATALINK column. However, a user cannot update the file while the file is currently linked. In order to modify the file, the user has to unlink the file in one transaction, modify the content and link the file in another transaction. So between the two transactions, the file is not linked, which means it is not under the control of the database (DataLinks) and the file reference is not visible in the database.
The database 80 stores an employee table 82 including a name column 83, a department column 84, and a picture column 81. The name column 83 typically contains a string, the department column an integer, while the picture column would contain a reference to an image stored in one of a number of external data files 20. The external data file 20 may also contain other forms of data, including documents, presentations, engineering drawings, spreadsheets, or video clips.
It would be undesirable for any one of the external data files 20 to get deleted, modified, or renamed while that external data file 20 continues to be referenced by the database 80, and in particular the picture column 81. A Datalinks File Manager (DLFM) 60 is therefore provided for maintaining a table (not illustrated) of linked file references. The DLFM 60 may also be responsible for controlling access permission and ownership of the external data files 20. The table maintained by the DLFM 60 contains attributes and subsets of the data stored in the external data files 20, along with logical references to the location of the external data files 20.
In operation, an application program 30 searches the database 80 via a SQL Application Programming Interface (API) request 40 to identify database entries 80 of interest. In the example illustrated in
The DBMS 50 is responsible for managing the links between the database 80 and the data files 20 and provides to the application 30 references to the selected external data files 20 via a handle. An example of the DBMS 50 is the DB2™ Universal Database product of International Business Machines Corporation. The application 30 can now access the external data file 20 directly using standard file-system API calls 70. Typical file-system API calls 70 are “file-open”, “file-read” and “file-delete”.
A Datalinks File System Filter DLFF 10 is a thin, database control layer or filter on the file system 75. The DLFF 10 intercepts certain of the file-system API calls 70 issued by the application 30. The file-system API calls 70 that are intercepted include file-open, file-rename, and file-delete calls. If the external data file 20 is referenced in the database 80, the DLFF 10 is responsible for enforcing referential-integrity constraints and access-control requirements defined for the data file 20. The referential integrity constraints ensure that the reference to data files 20 remains valid as long as it is “linked” to the database 80. DLFF 10 also ensures that any access call 70 meets Database Management System (DBMS) 50 access control requirements. An example of SQL mediated access is in the use of an authorization token, which is generated by the DBMS as part of the handle and verified by DLFF.
The DLFF 10 validates any authorization token (not illustrated) embedded in the file pathname for a ‘file-open’ operation. For example, when a user of application 30 submits a SQL API Request 40 to retrieve the employee picture (image) from the database 80, the DBMS 50 checks to see if the user has permission to access the employee table 82 containing the picture column 81. The permission may include select and view privileges on the employee table 82. The DBMS 50 returns the file name of the external file 20 to the application 30 only if the user has the required permission. An authorization token is embedded in the file name as part of the value returned for the picture column 81 by the DBMS 50. The application 30 then uses the file API call 70 to retrieve the image from the external file 20. When the DLFF 10 intercepts the file-open request, the DLFF 10 validates the authorization token to determine whether or not to pass the file-open request through to the native file system 75. If the validation fails, the file-open request is rejected. Once access has been authorized by a valid token, the application 30 interacts directly with the file system 75 for delivery of the external file 20 without the need for the DLFF 10 to further control the file access. This allows the application 30 the same speed of access as to a native file system.
In accordance with a first aspect of the invention, there is provided a method of maintaining consistency of content of an object and meta-data related to the object in a loose transaction model for object and meta-data updates. The method includes the steps of storing the related meta-data and a reference to the object in a table of a database, the object being stored externally to the database in an object store, the reference used to obtain a handle for directly accessing or manipulating the external object; obtaining a version number embedded in the handle and comparing the embedded version number with a version number of a latest committed version of the externally stored object to determine if the handle refers to a current version of the externally stored object.
Preferably the method further includes the steps of, if the encoded version number and the version number of the latest committed version match, comparing a last modification time stamp of the object with a last modification time stamp for the latest committed version of the object; and if the last modification time stamp of the object matches with the last modification time stamp for the latest committed version of the object, permitting access to the externally stored object. The method may further include the step of, if the last modification time stamp of the object does not match with the last modification time stamp for the latest committed version of the object, generating an error to indicate that the handle refers to stale content in the object.
Preferably the method further includes the steps of updating the object in-place under either DBMS control or file system control and linking the meta-data and the object under DBMS control.
Preferably the method further includes wherein the loose-transaction update model uses SQL Mediated Object Manipulation (SMOM) for an object that resides external to the database.
Preferably the method further includes the step of intercepting a native access to the externally stored object or a file system and validating the caller's access rights based on a combination of the version number and a last modification time stamp for a version of the object. The method may further include the intercepting step being carried out using a filter layer of the object store for the stored object.
Optionally, the object store is a local file system, or a distributed file system, the object being accessed from a remote file system client. A file access which occurs in the presence of authoritative caching and the comparing steps may be performed at the file system client. The method may further include the steps of caching the last known version number and the corresponding last modification time stamp at the file system client after an access and refreshing the last known version number and the corresponding last modification time stamp with latest values from a file server the next time one or both of the comparisons fail with the previously cached values, in which case the comparing steps are retried with refreshed values.
Preferably the method further includes the object which is a file.
Preferably the method further includes the version number associated with the object is embedded in an access token.
Preferably the method further includes the version number is temporally unique.
Preferably the method further includes the last-modification-time stamp attribute associated with the object is maintained by the object store.
In further aspects of the invention, an apparatus, a computer program, a computer program product, and a system for maintaining meta-data and file-data consistency in a loose transaction model of object and meta-data updates are provided.
A small number of embodiments of the invention are described hereinafter with reference to the drawings, in which:
A method, an apparatus, a computer program, a computer program product and a system for maintaining (or ensuring) meta-data and object-data consistency in a loose transaction model of object and meta-data updates, from the perspective of a reader application, are described hereinafter. Preferably, the object may be a file and an object store may be a file system. In the following description, numerous details are set forth. It will be apparent to one skilled in the art, however, that the present invention may be practised without these specific details. In other instances, well-known features are not described in detail so as not to obscure the present invention.
With the embodiments of the invention, allowing file content to be accessed and updated in-place by directly accessing the file system natively is more convenient. Besides, since content edits can be relatively long running, a loose transaction model for file and meta-data updates is useful where the file can be edited independently of the meta-data. A user can update a file while it is currently linked. In addition, the meta-data for that file can be accessed by other users while the file is in the process of being updated. The meta-data is always visible, even while the file is being update. However, access to the file may be denied. SQL mediated access can be extended to allow update-in-place. Specifically, the WRITE PERMISSION clause in
The embodiments of the invention address the noted problem in the context of meta-data maintained in a database table associated with an external file reference to content that is stored in a file system or an object store external to the database.
Generally, a file update takes longer than meta-data updates or typical database transactions. In a typical update, access to the file is requested and the file is locked. The content of the file is modified as well as corresponding meta-data generated. The file is then unlocked (i.e., the file update is declared to be completed) and the meta-data is imported into the database. Changes to the content of the file happen outside a regular DB2 transaction.
In particular, the problem addressed by the embodiments of the invention is ensuring that a reader application always gets a consistent view of the content of a file and the meta-data related to the file, given a loose transaction model for file and meta-data updates. An error can acceptably be returned in situations where the file data being accessed is not consistent with the corresponding meta-data seen by the reader.
Solutions are provided by the embodiments of the invention, firstly for files accessed from a local file system 700. This is then extended to a distributed file system environment 800 for files accessed from remote file system clients in the presence of authoritative caching.
The file server 880 includes a local repository 882 and a file manager daemon 884. A control path 840 bi-directionally extends between the DBMS 824 of the meta-data server 820 and the file manager daemon 884 of the file server 880. Likewise, the file manager daemon 884 of the file server 880 is bi-directionally coupled with the proxy daemon 890 of the client 830. The file server 886 also includes a distributed file system server 886. The distributed file system server 886 is coupled to a file server store 888. The distributed file system server 886 of the file server 880 is bi-directionally coupled (indicated by dashed line) to the distributed file system client 854 of the client 830.
The file is updated using normal file system application program interfaces (API's) 732, 832. The file can be updated via SQL Mediated Object Manipulation (SMOM) where a handle is obtained from the mediator 724, 824, that is preferably a filename with an encrypted access token string embedded as part of the filename, and is supplied as the filename to the filesystem API. The file may be updated several times before the file's meta-data is updated in the database 728, 828. To effect the corresponding meta-data updates and relate those updates to the file updates, the user issues an SQL update through the mediator DBMS 724, 824, optionally passing in the object handle, and commits both file and meta-data updates together. Another possibility is that the file may be updated based on the filesystem permissions. To effect the corresponding meta-data updates and relate those updates to the file updates, the user issues an SQL update through the mediator DBMS 724, 824, and commits both file and meta-data updates together.
Preferably, a filter layer 752, 852 transparently intercepts native access to the external object store or file system 756, 856 and 888, making use of the embodiments of the invention, to allow or deny access to the object based on a combination of an implicit version number and the last modification timestamp for a given version. This is done to validate the consistency of the meta-data retrieved by the reader and the content of the file.
The file edits happen outside of the DB transaction. The file updates get committed to the database (DB) via the DBMS 724, 824 once the update is done. DBMS 724, 824 communicates with the file management mediator daemon (also referred to as DLFM in the Datalinks product) 736, 884 passing a version number or generation number associated with the file, which is temporally unique on every committed update to the file. The file management mediator daemon 736, 884 utilizes a local repository (i.e. local to the fileserver node) 734, 882 to store the version number attribute of the file. (The repository 734, 882 could be a local database, as is the case of the embodiment implemented as part of the DB2 Datalinks product). The file management mediator daemon 736, 884 also saves a last modification timestamp attribute of the file maintained natively by the file system (or external store) as the last modified time of the latest committed version of the file.
This file management mediator daemon 736, 884 utilizes the local repository 734, 882 (i.e. local to the fileserver node) to store the attributes of these objects at the time of their association with the DB, and at the time of committed updates. (Again, the repository could be a local database, as is the case of the embodiment implemented as part of the DB2 Datalinks product)
The solution provided by the embodiments of the invention extends the functionality of this filter layer 752, 852 and the associated file management mediator daemon 736, 840 to ensure file-data meta-data consistency. The solution returns an error for file accesses where the file content are no longer consistent with the meta-data associated with the handle being used to reference the object. Preferably, the error code is ESTALE in the Unix operating system.
It may be observed that the token (embedded in the handle used to reference the object) becomes obsolete in the following situations:
These uncommitted file updates may have happened after the token is generated, or may have been pending from before the token generation. In this case, the meta-data in the table at the time the token was generated corresponds to the last committed file data and does not reflect uncommitted file updates that happened before the token was generated.
Such uncommitted file updates can happen via a re-linking of the same file (after an unlink and a change in content via a rename or a write to the file), or through an SQL Update commit for an update-in-place. There are two cases here:
The access token preferably contains a one-way hash value, h which is computed as follows:
h=hash(K, Tx, file name, servername, token length, flags),
where K is a secret symmetric key and Tx is the expiration time and the flags contain information about the token including which type of access is granted. Tx, token length and flags are also passed as part of the token. For cases where consistency is to be maintained between the DBMS meta-data and the object, the token can be setup to contain a version number and a different one-way hash value, H, computed as:
H=hash(K, Tx, file name, server name, token length, flags, Vc),
where Vc is the latest committed version at which the token is created and flags can contain information indicating that consistency is requested.
On the file manager side 736 for token validation, the file system filter 752 passes the token up to the file manager daemon 736. If the flags indicate that consistency is requested, the daemon 736 validates the token by calculating H where Vc is obtained from the last committed version number stored in the daemon's repository tables. The daemon 736 checks that the token has not expired and compares H with the one way hash contained within the access token.
Steps 614 and 616 in the previous section may be modified to just use the one-way hash value as the basis for checking for a version number match, instead of requiring that the version number be extracted from the token/handle. Step 614 could instead just retrieve the hash function portion of the token, H(token) and compute the one-way hash value H(commit) corresponding to V(commit), while the comparison for detecting staleness in step 616 may be replaced by “if H(token)≠H(commit)”.
A content management system is described with reference to
The table 204 may contain more than one DATALINK column. The entry for the first row has “1000” in column 1 and URL' in the DATALINK column DL1. The second row has “1019” in column 1 and NULL in column DL1. As indicated by a dotted line, URL' in row 1 of the table 204 refers to file 252 in the file system 250. The file 252 is stored externally to the database 202, and the content of the file 252 may include electronic documents, text, XML and HTML, digital audio and video, scanned images, facsimiles. The foregoing are merely examples of the type of the file 252. Other file types may be practiced without departing from the scope and spirit of the invention.
The content management system 200 also includes a DB2 agent 220 coupled to the database 202, a Data Link File Manager (DLFM) 230, and a Data Link File System Filter (DLFF) 240. An application 210 is a reader in the system 200 of FIG. 2.
The operation of the content management system 200 is described with reference to the flow diagrams of
With reference to
Arrow 264 extends from DB2 agent 220 to application 210. In step 264, a write access token is returned. The write access token includes a hash value that encodes the latest version number. Arrow 266 extends between the application 210 and DLFF 240. The application 210 opens the file for a write operation using a write access token via the DLFF 240. Arrow 268 extends from the DLFF 240 to the DLFM 230. The DLFF 240 forwards the write access token to the DLFM 230 for token validation. If the token is valid, then DLFM also checks if the file-data is consistent with the meta-data corresponding to the token. Arrow 270 extends from the DLFF module 240 to the application 210. In step 270 of
Arrow 272 extends from the application 210 to the file 252 in the file system 250. In step 272, the application 210 modifies the file 252 using the file descriptor. Arrow 274 extends from the application 210 to the DB2 agent module 220. In step 274, once the file update is completed, the application 210 issues a SQL UPDATE to the database DB2 202 via the DB2 agent 220 and imports other meta-data into the database. Arrow 276 extends from the DB2 agent 220 to the DLFM 230 in FIG. 2A. In step 276, the DB2 agent 220 invokes a DB2-DLFM update API and update the DLFM meta-data. At this stage, the DLFM determines the last modification timestamp of the file (same as the Tm(commit)) and records that timestamp in its local repository along with the updated version number (same as the link recovery id).
With reference to
In the case of a read operation, the flow is as follows. The application 210 requests a read access token from the DB2 agent 220 via a SQL SELECT statement. A read access token is returned. The read access token includes a hash value that encodes the latest version number. The application 210 opens the file for a read operation using a read access token via the DLFF 240. The DLFF 240 forwards the read access token to the DLFM 230 for token validation. If the token is valid, then DLFM also checks if the file-data is consistent with the meta-data corresponding to the token. This is how the DLFM performs the check. The DLFM looks up the latest committed version number V(commit) and the corresponding modification timestamp Tm(commit) in its local repository (R). Then the DLFM determines if the version number embedded in the token, i.e. V(token) matches V(commit) (preferably by computing the one-way hash value corresponding to V(commit) and comparing it with the hash value embedded in the token). In case of a mismatch, the DLFM returns an error to DLFF. Otherwise, the DLFM obtains the last modification timestamp of the file Tm(access) from the filesystem, and determines if Tm(access) is later than Tm(commit). If so, then the DLFM has detected an inconsistency and returns an error to DLFF, which reports ESTALE to the application. If not, then the DLFM returns OK and DLFF returns success. The file system returns the file descriptor to the application 210 for a successful open. The application 210 reads the file 252 using the file descriptor.
2.3 Extension to a Distributed File System Environment
A further embodiment of the invention extends the above solution to a distributed environment 800 of
A proxy daemon 890 may be set up at each client 830 to service upcalls 838 from the filter module 852, which intercepts file accesses on the client 830. This proxy daemon 890 contacts the central server 880/repository 882 when required. As described hereinbefore, the requirement is to minimize these contacts, especially for read performance in the most typical cases (i.e. repeated accesses to the same file should be servicable from the cache without contacting the server as far as possible).
First Time Access
The first time a file is accessed with a token at the client, the client proxy daemon 890 contacts the remote central repository 882 and makes an entry in a local (possibly in-memory) table with:
The check for stale data is:
Cached Case (Repeated Accesses):
The next time the same file is accessed (either with the same or a different token), the proxy daemon 890 can simply look up the proxy daemon's local table entry and if Fid(access)=Fid(commit), then the proxy daemon 890 can use the in-memory entry to perform the above check.
Fid(commit) and Fid(access) refer to the file-id value at the time of commit and time of access respectively, i.e. Fid(commit)=value of Fid (Filename, Tm(commit)) where Fid(f,t)=value of unique file id of the file named “f” at time “t”.
The reason for explicitly introducing these in the distributed case, is to correctly handle cases where the file may have been unlinked, renamed and relinked under the same name (as in Scenario 2). By indexing V(commit) and Tm(commit) against Fid(commit) in the local in-memory table, and looking up the cached entry against Fid(access). This ensures that the version number and Tm values refer to the same file, so that a comparison with Tm(access) makes sense. This is important in the distributed case, because the cached value of V may be stale, making Tm(access) the real determinant of whether the entry should be considered stale or not.
The unique file id maintained by a distributed filesystem usually includes some kind of per inode generation number of its own to account for reuse of inode numbers after files are deleted. This inode generation number is typically incremented every time an inode number is reassigned to a newly created file. This ensures the temporal uniqueness of a file identifier.
Just as in the local case, if Tm(access)>Tm(commit) and the token is a write token, an optional step of checking the file if an update-in-progress with the same token/user may be involved (e.g. by checking state information in the repository).
This approach causes the client to contact the central repository on the server only if the file changed or got relinked since the last time the entry was looked up. The in-memory table entries can be flushed/reused based on the expiration time and resource limits.
In step 668, the current value of the timestamp Tm(access) is obtained from the file. In decision block 670, a check is made to determine if the timestamp Tm(commit) is equal to Tm(access). If decision block 670 returns false (NO), processing continues at decision block 672. If decision block 670 returns true (YES), the file and meta-data are consistent, the variable refreshed is set to false, and access is allowed to the object in step 622.
In decision block 672, a check is made to determine if the cache was accessed (i.e. REFRESHED equals false). If decision block 672 returns true (YES), processing continues at block 660. Finally, if decision block 672 returns false (NO), the meta-data and file data are inconsistent and an error is returned in step 624.
This section discusses various key scenarios and how the embodiments handle each of these situations to prevent potential inconsistencies from arising.
It is sufficient to guarantee the consistency at the point of the file/object open. The user is expected to use some form of file/application-level locking as in a normal file system to ensure serialization between concurrent readers and updaters. Specifically, an updater must not start its updates if there is any reader that has opened the file. Otherwise, the reader could see inconsistent data. Table 1 is a legend of terminology used hereinafter.
TABLE 1 Term Meaning Qr(f, t) user does a SQL query and an access token for a file f is assigned at time t. Qw(f, t) user does the SQL query for a write token and the write token for the file f is assigned at time t. C(f, t) Content of the file f at time t. Rs(f, t) time when the reader opens the file f with access token for generated at t. Re(f, t) time when the reader closes the file f (finished reading the file) corresponding to Rs(f, t). Us(f, t) time when the writer opens the file f for update using the write token generated at t. Ue(f, t) time when the writer closes the file f (i.e. finishes writing the file) corresponding to Us(f, t), but not yet committed to the database (DB) using SQL update. Uc(f, t) time when meta-data in DB is updated and committed together with the file-data updates for the file f. I(f) time when the file f is linked (inserted into) to the database. D(f) time when the file f is unlinked (deleted) from the database (restore on unlink). Us(f) time when the writer opens the file for an update without a token. Ue(f) time when the writer closes the file (finishes writing the file) corresponding to Us(f). M(f′, f) time when the file f′ is moved to f (could be rm f, mv f′ f) results in f getting replaced by f′ − both data + attributes (including timestamp) C(f, ta) <= C(f′, tb), where ta < M(f′, f) < tb.
3.1 Addressing Potential Inconsistency Scenarios
The application is responsible for ensuring serialization, so that Rs-Re essentially happens under a shared file lock and Us-Ue happens under an exclusive file lock. However, multiple Us-Ue iterations can happen before Uc to finally commit the update, and an Rs-Re can happen between two such Us-Ue sequences.
As in scenario (1) above, with the embodiments of the invention, Rs(f,t) fails since V(token)<V(commit), i.e., the version-id embedded in the read-token is earlier than the one in I2(f). In the distributed file system case, where the client's in memory table has a stale value of V(commit) and Tm(commit), the condition [Tm(access) for f′>Tm(commit) for f] may not suffice to detect stale information. However, the check Fid(access)=Fid(commit) fails, and a refresh of the values of V(commit) and Tm(commit) from the server is triggered.
The file data consistent with Qr(f,t2) is C(f,I(f)) or C(f,t1), but not C(f,t2) since the meta-data has not changed yet at t2 but only after that. C(f,Rs(f,t2)) differs from C(f, t1) as the file content has been updated between t1 and t2.
With the embodiments of the invention, Rs(f,t2) fails since V(token)<V(commit).
Another case that can be considered equivalent to the above for current purposes is:
I(f); Qw(f,t1); Us(f,t1); Qr(f,t2); Ue(f,t2); Uc(f,t1); Rs(f,t2).
With the embodiments of the invention, Rs(f,t2) fails since V(token)<V(commit).
With the embodiments of the invention, Rs (f,t2) fails since V(token)<V(commit).
As the foregoing embodiments of the invention illustrate, a loose transaction model for updates to a file and its corresponding meta-data through a mediator is useful for directly performing in-place edits of content data residing on stores external to the indexed meta-data store (the latter could be a DBMS). This is subject to the requirement of ensuring consistency between the file content and the associated meta-data from a reader's perspective. The embodiments of the invention encode a version number in the handle for referencing the object associated with a given meta-data state. A thin interceptor layer on the native store, where the objects content are stored, decodes this version number and compares the decoded version number with the version number of the latest committed version of the file. This is done to determine if the handle refers to the current version. If the version matches, the thin interceptor layer checks for uncommitted updates by comparing the last modification time stamp of the file with the last modification timestamp for the latest committed version. If these match, the thin interceptor layer allows access to proceed as usual for the file. Otherwise, the thin interceptor layer reports an error indicating that the handle refers to stale data.
The embodiments of the invention are advantageous in a number of ways. The embodiments enable in-place updates of file content to be made directly on the file system, which is separate from the meta-data store. This is done while ensuring that a reader application does not see an inconsistency between the meta-data and the file data. Another advantage is that the consistency check on file access can be performed without contacting the meta-data server, which is separate from the meta-data store. This results in fast content access times even when the file and meta-data servers are distributed.
Advantageously, this approach is suitable for a distributed model for file and meta-data storage, since the file content access path checks do not require any direct communication with the meta-data server. This approach extends easily to a configuration where the external store is a distributed file system and the content is accessed directly from distributed file system clients. This works even in the presence of authoritative caching (as in DCE-DFS) with minimal communication overheads and does so using an entirely client initiated approach, without any client specific state being required to be maintained on the servers. The approach may be enhanced to work correctly (in terms of preventing accesses to potentially inconsistent data) with mobile file systems (e.g., Coda or Intermezzo) that operate in disconnected mode. The consistency check makes efficient use of caching for improved performance without losing correctness.
Further, the method described does not require clock synchronization between the database server and the filesystem server, and even with the filesystem client in the distributed filesystem case. Note that even in the scenario where the clocks are synchronized, it is not efficient for the DB server to save the time when the latest update was committed, T1(Commit) (instead of V(commit)), and provide it to the mediator daemon at transaction commit to save the same (T1(commit) value) for the corresponding files in the local repository on the file server. The reason is that this would require the database server to track all the rows involved in the linkfile operation for a particular transaction, update them prior to doing its commit processing with the T(Commit) value, and the same would have to be done at the file management daemon's end.
The solution is capable of covering cases where the database has been rolled back to an earlier state, as might happen in the case of a point-in-time restore, and should also work with database replication, as long as the legitimacy of the last modified timestamp of the file is maintained during a restore and across replicas.
The embodiments of the invention are preferably implemented using one or more general-purpose computers. In particular, the processing or functionality of
In particular, the software may be stored in a computer readable medium, including the storage devices described below. The software is preferably loaded into the computer from the computer readable medium and then carried out by the computer. A computer program product includes a computer readable medium having such software or a computer program recorded on it that can be carried out by a computer. The use of the computer program product in the computer preferably effects advantageous apparatuses for maintaining (or ensuring) meta-data and file-data consistency in a loose transaction model of file and meta-data updates.
Preferably, a computer system includes a computer, a video display, and input devices. In addition, the computer system can have any of a number of other output devices including line printers, laser printers, plotters, and other reproduction devices connected to the computer. The computer system can be connected to one or more other computers via a communication interface using an appropriate communication channel such as a modem communications path, a computer network, or the like. The computer network 820 may include a local area network (LAN), a wide area network (WAN), an Intranet, and/or the Internet.
The computer itself consists of a central processing unit(s) (simply referred to as a processor hereinafter), a memory which may include random access memory (RAM) and read-only memory (ROM), input/output (IO) interfaces, a video interface, and one or more storage devices. The storage device(s) can consist of one or more of the following: a floppy disc, a hard disc drive, a magneto-optical disc drive, CD-ROM, magnetic tape or any other of a number of non-volatile storage devices well known to those skilled in the art. Each of the components is typically connected to one or more of the other devices via a bus that in turn can consist of data, address, and control buses.
The foregoing system is simply provided for illustrative purposes and other configurations can be employed without departing from the scope and spirit of the invention. Computers with which the embodiment can be practiced include IBM-PC/ATs or compatibles, one of the Macintosh (TM) family of PCs, Sun Sparcstation (TM), a workstation or the like. The foregoing are merely examples of the types of computers with which the embodiments of the invention may be practiced. Typically, the processes of the embodiments, are resident as software or a program recorded on a hard disk drive as the computer readable medium, and read and controlled using the processor. Intermediate storage of the program and intermediate data and any data fetched from the network may be accomplished using the semiconductor memory, possibly in concert with the hard disk drive.
In some instances, the program may be supplied to the user encoded on a CD-ROM or a floppy disk, or alternatively could be read by the user from the network via a modem device connected to the computer, for example. Still further, the software can also be loaded into the computer system from other computer readable medium including magnetic tape, a ROM or integrated circuit, a magneto-optical disk, a radio or infra-red transmission channel between the computer and another device, a computer readable card such as a PCMCIA card, and the Internet and Intranets including email transmissions and information recorded on websites and the like. The foregoing are merely examples of relevant computer readable mediums. Other computer readable mediums may be practiced without departing from the scope and spirit of the invention.
In the foregoing manner, a method, an apparatus, a computer program, a computer program product, and a system for maintaining (or ensuring) meta-data and file-data consistency in a loose transaction model of file and meta-data updates are disclosed. While only a small number of embodiments are described, it will be apparent to those skilled in the art in view of this disclosure that numerous changes and/or modifications can be made without departing from the scope and spirit of the invention.
|Cited Patent||Filing date||Publication date||Applicant||Title|
|US5499365 *||Aug 4, 1993||Mar 12, 1996||International Business Machines Corporation||System and method for controlling versions of objects in an object oriented computing environment|
|US6029160||Aug 15, 1997||Feb 22, 2000||International Business Machines Corporation||Method and means for linking a database system with a system for filing data|
|US6088694 *||Mar 31, 1998||Jul 11, 2000||International Business Machines Corporation||Continuous availability and efficient backup for externally referenced objects|
|US6205445 *||Mar 24, 1999||Mar 20, 2001||Matsushita Electric Industrial Co., Ltd.||File distribution system and method thereof|
|US6564215 *||Dec 16, 1999||May 13, 2003||International Business Machines Corporation||Update support in database content management|
|US6581075 *||Dec 28, 2000||Jun 17, 2003||Nortel Networks Limited||System and method for database synchronization|
|1||"Filter File System Implementation of Distributed File Systems", U.S. Appl. No. 09/571,778, IBM, filed May 16, 2000), 30 pages.|
|2||ISO/IEC 9075-9:2000(E), "Information Technology-Database Languages SL-Part 9: Management of External Data (SQL/MED)".|
|Citing Patent||Filing date||Publication date||Applicant||Title|
|US7020805 *||Aug 15, 2002||Mar 28, 2006||Sun Microsystems, Inc.||Efficient mechanisms for detecting phantom write errors|
|US7051155||Aug 5, 2002||May 23, 2006||Sun Microsystems, Inc.||Method and system for striping data to accommodate integrity metadata|
|US7103811||Dec 23, 2002||Sep 5, 2006||Sun Microsystems, Inc||Mechanisms for detecting silent errors in streaming media devices|
|US7133883||Dec 23, 2002||Nov 7, 2006||Sun Microsystems, Inc.||General techniques for diagnosing data corruptions|
|US7222117 *||Nov 14, 2003||May 22, 2007||Advent Software, Inc.||Segmented global area database|
|US7320076 *||Mar 5, 2003||Jan 15, 2008||Sun Microsystems, Inc.||Method and apparatus for a transaction-based secure storage file system|
|US7328374 *||Apr 29, 2004||Feb 5, 2008||International Business Machines Corporation||Method and apparatus for implementing assertions in hardware|
|US7353387 *||Mar 8, 2001||Apr 1, 2008||International Business Machines Corporation||Method and system for integrating encryption functionality into a database system|
|US7353432||Nov 25, 2003||Apr 1, 2008||Sun Microsystems, Inc.||Maintaining high data integrity|
|US7500020||Dec 31, 2003||Mar 3, 2009||Symantec Operating Corporation||Coherency of replicas for a distributed file sharing system|
|US7624119 *||Feb 11, 2004||Nov 24, 2009||International Business Machines Corporation||Low-overhead built-in timestamp column for relational database systems|
|US7647360 *||Jun 19, 2006||Jan 12, 2010||Hitachi, Ltd.||System and method for managing a consistency among volumes in a continuous data protection environment|
|US7734820||Dec 31, 2003||Jun 8, 2010||Symantec Operating Corporation||Adaptive caching for a distributed file sharing system|
|US7743023 *||Feb 1, 2006||Jun 22, 2010||Microsoft Corporation||Scalable file replication and web-based access|
|US7809778 *||Mar 8, 2006||Oct 5, 2010||Omneon Video Networks||Idempotent journal mechanism for file system|
|US7926066 *||Apr 5, 2005||Apr 12, 2011||Openpages, Inc.||Adaptive content platform and application integration with the platform|
|US7971144||Oct 31, 2007||Jun 28, 2011||Openpages||Adaptive content platform and method of using same|
|US8171003||Jan 15, 2010||May 1, 2012||Kunio Kamimura||Method and apparatus for changing reference of database|
|US8291269 *||Nov 14, 2011||Oct 16, 2012||Advent Software, Inc.||Multi-writer in-memory non-copying database (MIND) system and method|
|US8332349||Jan 6, 2012||Dec 11, 2012||Advent Software, Inc.||Asynchronous acid event-driven data processing using audit trail tools for transaction systems|
|US8458217||Aug 24, 2009||Jun 4, 2013||Advent Software, Inc.||Instantly built information space (IBIS)|
|US8495658||Mar 30, 2011||Jul 23, 2013||International Business Machines Corporation||Adaptive content platform and application integration with the platform|
|US8504758||Sep 21, 2010||Aug 6, 2013||Amazon Technologies, Inc.||System and method for logical deletion of stored data objects|
|US8572048 *||Nov 10, 2009||Oct 29, 2013||TIM LaBERGE||Supporting internal consistency checking with consistency coded journal file entries|
|US8589957||Jun 27, 2011||Nov 19, 2013||International Business Machines Corporation||Adaptive platform|
|US8615531 *||Sep 29, 2008||Dec 24, 2013||Xcerion Aktiebolag||Programmatic data manipulation|
|US8620863 *||Sep 29, 2008||Dec 31, 2013||Xcerion Aktiebolag||Message passing in a collaborative environment|
|US8650156 *||Dec 23, 2010||Feb 11, 2014||Amazon Technologies, Inc.||System and method for fetching the latest versions of stored data objects|
|US8671339||Jan 7, 2011||Mar 11, 2014||Open Text S.A.||System, method and computer program product for asset sharing among hierarchically interconnected objects|
|US8689207 *||Aug 4, 2009||Apr 1, 2014||Red Bend Ltd.||Performing an in-place update of an operating storage device|
|US8769350 *||Sep 20, 2011||Jul 1, 2014||Advent Software, Inc.||Multi-writer in-memory non-copying database (MIND) system and method|
|US8805878 *||Oct 8, 2003||Aug 12, 2014||Open Text S.A.||System and method for managing enterprise-level interrelated site, channel, and content objects|
|US8849825||Dec 23, 2010||Sep 30, 2014||Amazon Technologies, Inc.||System and method for clustering distributed hash table entries|
|US8886671||Aug 14, 2013||Nov 11, 2014||Advent Software, Inc.||Multi-tenant in-memory database (MUTED) system and method|
|US8959123 *||Sep 29, 2008||Feb 17, 2015||Xcerion Aktiebolag||User interface framework|
|US8996831||Jul 29, 2013||Mar 31, 2015||Amazon Technologies, Inc.||System and method for logical deletion of stored data objects|
|US9047344||Oct 17, 2005||Jun 2, 2015||International Business Machines Corporation||Guaranteeing data and metadata referential integrity in content management archival solutions|
|US9053054||Feb 10, 2014||Jun 9, 2015||Amazon Technologies, Inc.||System and method for fetching the latest versions of stored data objects|
|US9071623 *||Sep 29, 2008||Jun 30, 2015||Xcerion Aktiebolag||Real-time data sharing|
|US9235476||Mar 30, 2015||Jan 12, 2016||Amazon Technologies, Inc.||System and method for logical deletion of stored data objects|
|US9344497||Sep 29, 2008||May 17, 2016||Xcerion Aktiebolag||State management of applications and data|
|US9516073||Dec 30, 2013||Dec 6, 2016||Open Text Sa Ulc||System, method and computer program product for asset sharing among hierarchically interconnected objects|
|US20020129260 *||Mar 8, 2001||Sep 12, 2002||Bruce Benfield||Method and system for integrating encryption functionality into a database system|
|US20040034817 *||Aug 15, 2002||Feb 19, 2004||Talagala Nisha D.||Efficient mechanisms for detecting phantom write errors|
|US20040123032 *||Dec 24, 2002||Jun 24, 2004||Talagala Nisha D.||Method for storing integrity metadata in redundant data layouts|
|US20040133539 *||Dec 23, 2002||Jul 8, 2004||Talagala Nisha D||General techniques for diagnosing data corruptions|
|US20040153746 *||Jul 31, 2003||Aug 5, 2004||Talagala Nisha D.||Mechanisms for embedding and using integrity metadata|
|US20040175000 *||Mar 5, 2003||Sep 9, 2004||Germano Caronni||Method and apparatus for a transaction-based secure storage file system|
|US20050060553 *||Oct 28, 2004||Mar 17, 2005||Microsoft Corporation||Session-state manager|
|US20050177590 *||Feb 11, 2004||Aug 11, 2005||Chen Yao-Ching S.||Low-overhead built-in timestamp column for relational database systems|
|US20050251707 *||Apr 29, 2004||Nov 10, 2005||International Business Machines Corporation||Mothod and apparatus for implementing assertions in hardware|
|US20050289532 *||Apr 5, 2005||Dec 29, 2005||Openpages Inc.||Adaptive content platform and application integration with the platform|
|US20070023507 *||Jul 26, 2005||Feb 1, 2007||Microsoft Corporation||Contact data structure and management|
|US20070088754 *||Oct 17, 2005||Apr 19, 2007||International Business Machines Corporation||Method and system to guarantee data and metadata referential integrity in content management archival solutions|
|US20070185932 *||Feb 1, 2006||Aug 9, 2007||Microsoft Corporation||Scalable File Replication and Web-based Access|
|US20070214195 *||Mar 8, 2006||Sep 13, 2007||Omneon Video Networks||Idempotent journal mechanism for file system|
|US20070294274 *||Jun 19, 2006||Dec 20, 2007||Hitachi, Ltd.||System and method for managing a consistency among volumes in a continuous data protection environment|
|US20080052729 *||Oct 31, 2007||Feb 28, 2008||Santanu Paul||Adaptive content platform and method of using same|
|US20090070304 *||Nov 17, 2008||Mar 12, 2009||International Business Machines Corporation||Low-overhead built-in timestamp column for relational database systems|
|US20090164592 *||Sep 29, 2008||Jun 25, 2009||Xcerion Ab||Network operating system|
|US20090171974 *||Sep 29, 2008||Jul 2, 2009||Xcerion Ab||Network operating system|
|US20090172087 *||Sep 29, 2008||Jul 2, 2009||Xcerion Ab||Network operating system|
|US20090177734 *||Sep 29, 2008||Jul 9, 2009||Xcerion Ab||Network operating system|
|US20100031245 *||Aug 4, 2009||Feb 4, 2010||Red Bend Ltd.||Performing an In-Place Update Of an Operating Storage Device|
|US20100131479 *||Jan 15, 2010||May 27, 2010||Athena Telecom Lab, Inc.||Method and apparatus for changing reference of database|
|US20100198789 *||Jun 4, 2008||Aug 5, 2010||Kunio Kamimura||Database contradiction solution method|
|US20110082833 *||Jun 3, 2009||Apr 7, 2011||Kunio Kamimura||Database parallel editing method|
|US20110106888 *||Jan 7, 2011||May 5, 2011||Vignette Software Llc||System, method and computer program product for asset sharing among hierarchically interconnected objects|
|US20110113017 *||Nov 10, 2009||May 12, 2011||Quantum Corporation||Supporting Internal Consistency Checking with Consistency Coded Journal File Entries|
|US20110137862 *||Feb 10, 2011||Jun 9, 2011||Athena Telecom Lab, Inc.||Method and apparatus for parallel edit to editable objects|
|US20110179425 *||Mar 30, 2011||Jul 21, 2011||Openpages, Inc.||Adaptive Content Platform and Application Integration with the Platform|
|US20130346810 *||Jun 14, 2013||Dec 26, 2013||Netapp. Inc.||Use of application-level context information to detect corrupted data in a storage system|
|US20140181030 *||Feb 28, 2014||Jun 26, 2014||Metaswitch Networks Ltd||Identifying data items|
|US20140304388 *||Jun 20, 2014||Oct 9, 2014||Open Text SA||Site and channel management to achieve content repurposing|
|US20160196262 *||Feb 24, 2016||Jul 7, 2016||International Business Machines Corporation||Method and system for backup and recovery|
|U.S. Classification||1/1, 707/999.203, 707/999.008, 707/999.01|
|International Classification||G06F17/30, G06F12/00|
|Cooperative Classification||Y10S707/99938, Y10S707/99954, G06F8/71, G06F17/30943, G06F17/30551|
|European Classification||G06F17/30S4P8T, G06F17/30Z|
|Jan 3, 2002||AS||Assignment|
Owner name: INTERNATIONAL BUSINESS MACHINES CORPORATION, NEW Y
Free format text: ASSIGNMENT OF ASSIGNORS INTEREST;ASSIGNORS:NARANG, INDERPAL SINGH;BRANNON, KAREN WOLFE;BHATTACHARYA, SUPARNA;AND OTHERS;REEL/FRAME:012434/0837;SIGNING DATES FROM 20010928 TO 20011016
|Sep 9, 2008||FPAY||Fee payment|
Year of fee payment: 4
|Nov 12, 2012||REMI||Maintenance fee reminder mailed|
|Jan 28, 2013||FPAY||Fee payment|
Year of fee payment: 8
|Jan 28, 2013||SULP||Surcharge for late payment|
Year of fee payment: 7
|Jul 18, 2016||FPAY||Fee payment|
Year of fee payment: 12